On Fri, 2 Jun 2023 23:31:52 -0700, Jakub Kicinski <kuba at kernel.org> wrote:> On Fri, 2 Jun 2023 17:22:06 +0800 Xuan Zhuo wrote: > > drivers/net/virtio_net.c | 163 +++++++++++++++++++++++++++++++++------ > > ack for this going via the vhost tree, FWIW, but you'll potentially > need to wait for the merge window to move forward with the actual > af xdp patches, in this case.My current plan is to let virtio support premapped dma first, and then implement virtio-net to support af-xdp zerocopy. This will indeed involve two branches. But most of the implementations in this patch are virtio code, so I think it would be more appropriate to commit to vhost. Do you have any good ideas? Thanks.
Michael S. Tsirkin
2023-Jun-05 05:44 UTC
[PATCH vhost v10 10/10] virtio_net: support dma premapped
On Mon, Jun 05, 2023 at 10:10:44AM +0800, Xuan Zhuo wrote:> On Fri, 2 Jun 2023 23:31:52 -0700, Jakub Kicinski <kuba at kernel.org> wrote: > > On Fri, 2 Jun 2023 17:22:06 +0800 Xuan Zhuo wrote: > > > drivers/net/virtio_net.c | 163 +++++++++++++++++++++++++++++++++------ > > > > ack for this going via the vhost tree, FWIW, but you'll potentially > > need to wait for the merge window to move forward with the actual > > af xdp patches, in this case. > > > My current plan is to let virtio support premapped dma first, and then implement > virtio-net to support af-xdp zerocopy. > > This will indeed involve two branches. But most of the implementations in this > patch are virtio code, so I think it would be more appropriate to commit to > vhost. Do you have any good ideas? > > > Thanks.Are you still making changes to net core? DMA core? If it's only virtio-net then I can probably merge all of it - just a couple of bugfixes there so far, it shouldn't cause complex conflicts. -- MST